Improvement of Text Feature Selection Method Based on TFIDF

Abstract

TFIDF is a kind of common methods used to select the text feature, but it has many disadvantages. First, the method undervalues that this term can represent the characteristic of the documents of this class if it only frequently appears in the documents belongs to the same class while infrequently in the documents of the other class. Second TFIDF neglects the relations between the feature and the class. The paper proposed the improved TFIDF strategy, and combined with the text classification method of simple distance vector to compare to traditional TFIDF, and obtained the very good classified effect, the experiment proved its feasibility.